Upper-body kinesthetic teaching of a free-standing humanoid robot

Abstract

Abstract-We present an integrated approach allowing a free-standing humanoid robot to acquire new motor skills by kinesthetic teaching. The proposed method controls simultaneously the upper and lower body of the robot with different control strategies. Imitation learning is used for training the upper body of the humanoid robot via kinesthetic teaching, while at the same time Reaction Null Space method is used for keeping the balance of the robot. During demonstration, a force/torque sensor is used to record the exerted forces, and during reproduction, we use a hybrid position/force controller to apply the learned trajectories in terms of positions and forces to the end effector. The proposed method is tested on a 25-DOF Fujitsu HOAP-2 humanoid robot with a surface cleaning task.

Bipedal walking energy minimization by reinforcement learning with evolving policy parameterization

“…Machine learning has been successfully used before for learning tasks on bipedal robots, such as dynamic balancing, or even in tasks involving also the upper body such as learning to clean vertical surfaces [6]. One especially promising approach for autonomous robot learning is reinforcement learning (RL), as demonstrated in [7]- [11].…”

Reinforcement learning for improving imitated in-contact skills

“…Recently, researchers have shown interest in studying in-contact skills. In-contact tasks which have been learned from demonstration include cleaning a vertical surface [2], controlling stiffness [3], ball-in-box [4], pouring drink [4], box pulling [5], flipping task [5], stapling [6], and grasping small objects [7]. However, all these studies aim at learning a policy merely to reproduce the demonstrated force.…”

Learning dynamic graffiti strokes with a compliant robot

“…drawing workspace are computed by demonstrating a series of drawing movements by accompanying the manipulator along the drawing surface through kinesthetic teaching [33]. The trajectory to be tracked by the robot is then computed withx(t) = Ap(t).…”